Lower limb arthroplasty: can we produce a tool to predict outcome and failure, and is it cost-effective? An epidemiological study

Nigel Arden, Doug Altman, David Beard, Andrew Carr, Nicholas Clarke, Gary Collins, Cyrus Cooper, David Culliford, Antonella Delmestri, Stefanie Garden, Tinatin Griffin, Kassim Javaid, Andrew Judge, Jeremy Latham, Mark Mullee, David Murray, Emmanuel Ogundimu, Rafael Pinedo-Villanueva, Andrew Price, Daniel Prieto-AlhambraJames Raftery

Research output: Contribution to journalArticlepeer-review

49 Downloads (Pure)

Abstract

Background Although hip and knee arthroplasties are considered to be common elective cost-effective operations, up to one-quarter of patients are not satisfied with the operation. A number of risk factors for implant failure are known, but little is known about the predictors of patient-reported outcomes. Objectives (1) Describe current and future needs for lower limb arthroplasties in the UK; (2) describe important risk factors for poor surgery outcomes and combine them to produce predictive tools (for hip and knee separately) for poor outcomes; (3) produce a Markov model to enable a detailed health economic analysis of hip/knee arthroplasty, and for implementing the predictive tool; and (4) test the practicality of the prediction tools in a pragmatic prospective cohort of lower limb arthroplasty. Design The programme was arranged into four work packages. The first three work packages used the data from large existing data sets such as Clinical Practice Research Datalink, Hospital Episode Statistics and the National Joint Registry. Work package 4 established a pragmatic cohort of lower limb arthroplasty to test the practicality of the predictive tools developed within the programme. Results The estimated number of total knee replacements (TKRs) and total hip replacements (THRs) performed in the UK in 2015 was 85,019 and 72,418, respectively. Between 1991 and 2006, the estimated age-standardised rates (per 100,000 person-years) for a THR increased from 60.3 to 144.6 for women and from 35.8 to 88.6 for men. The rates for TKR increased from 42.5 to 138.7 for women and from 28.7 to 99.4 for men. The strongest predictors for poor outcomes were preoperative pain/function scores, deprivation, age, mental health score and radiographic variable pattern of joint space narrowing. We found a weak association between body mass index (BMI) and outcomes; however, increased BMI did increase the risk of revision surgery (a 5-kg/m2 rise in BMI increased THR revision risk by 10.4% and TKR revision risk by 7.7%). We also confirmed that osteoarthritis (OA) severity and migration pattern of the hip predicted patient-reported outcome measures. The hip predictive tool that we developed performed well, with a corrected R2 of 23.1% and had good calibration, with only slight overestimation of Oxford Hip Score in the lowest decile of outcome. The knee tool developed performed less well, with a corrected R2 of 20.2%; however, it had good calibration. The analysis was restricted by the relatively limited number of variables available in the extant data sets, something that could be addressed in future studies. We found that the use of bisphosphonates reduced the risk of revision knee and hip surgery by 46%. Hormone replacement therapy reduced the risk by 38%, if used for at least 6 months postoperatively. We found that an increased risk of postoperative fracture was prevented by bisphosphonate use. This result, being observational in nature, will require confirmation in a randomised controlled trial. The Markov model distinguished between outcome categories following primary and revision procedures. The resulting outcome prediction tool for THR and TKR reduced the number and proportion of unsatisfactory outcomes after the operation, saving NHS resources in the process. The highest savings per quality-adjusted life-year (QALY) forgone were reported from the oldest patient subgroups (men and women aged ≥ 80 years), with a reported incremental cost-effectiveness ratio of around £1200 saved per QALY forgone for THRs. In the prospective cohort of arthroplasty, the performance of the knee model was modest (R2 = 0.14) and that of the hip model poor (R2 = 0.04). However, the addition of the radiographic OA variable improved the performance of the hip model (R2 = 0.125 vs. 0.110) and high-sensitivity C-reactive protein improved the performance of the knee model (R2 = 0.230 vs. 0.216). These data will ideally need replication in an external cohort of a similar design. The data are not necessarily applicable to other health systems or countries. Conclusion The number of total hip and knee replacements will increase in the next decade. High BMI, although clinically insignificant, is associated with an increased risk of revision surgery and postoperative complications. Preoperative pain/function, the pattern of joint space narrowing, deprivation index and level of education were found to be the strongest predictors for THR. Bisphosphonates and hormone therapy proved to be beneficial for patients undergoing lower limb replacement. The addition of new predictors collected from the prospective cohort of arthroplasty slightly improved the performance of the predictive tools, suggesting that the potential improvements in both tools can be achieved using the plethora of extra variables from the validation cohort. Although currently it would not be cost-effective to implement the predictive tools in a health-care setting, we feel that the addition of extensive risk factors will improve the performances of the predictive tools as well as the Markov model, and will prove to be beneficial in terms of cost-effectiveness. Future analyses are under way and awaiting more promising provisional results. Future work Further research should focus on defining and predicting the most important outcome to the patient.
Original languageEnglish
Pages (from-to)1-246
JournalProgramme Grants for Applied Research
Volume5
Issue number12
DOIs
Publication statusPublished - 1 Jun 2017

Fingerprint

Dive into the research topics of 'Lower limb arthroplasty: can we produce a tool to predict outcome and failure, and is it cost-effective? An epidemiological study'. Together they form a unique fingerprint.

Cite this